Assessing Continuous Operator Workload With a Hybrid Scaffolded Neuroergonomic Modeling Approach

Area of Science:

• Neuroscience
• Human-Computer Interaction
• Machine Learning

Background:

• Adaptive systems need real-time mental workload assessment for dynamic task allocation.
• Neuroergonomic measures combined with task demand and performance data can clarify workload interpretation.
• Operator workload prediction is crucial for enhancing system performance and safety.

Purpose of the Study:

• To predict operator workload from neurological data using statistical learning.
• To develop neuroergonomic methods for operator state assessment in adaptive systems.
• To explore the efficacy of machine learning models for inferring workload from EEG data.

Main Methods:

• Utilized electroencephalogram (EEG) data as input for machine learning algorithms.
• Employed statistical learning to fit neurological-to-state-assessment models.
• Used simulation-generated workload profile data for model training and validation.

Main Results:

• Cross-participant models statistically distinguished between 62% of workload changes.
• Machine learning models trained on resampled workload profiles performed comparably to those trained on deterministic profiles.
• Stochastic models are viable for cross-participant workload prediction with limited training data.

Conclusions:

• Novel modeling techniques enable operator state assessment using neuroergonomic methods.
• Trained models can be directly applied to neurophysiological data for future assessments.
• These findings support the integration of neuroergonomic assessments into adaptive systems.